[EMAIL PROTECTED] wrote:

> Hello Brian,
>
> What about:
>     vos remove $server $failing_partition $volume.readonly
>     vos addsite $new_server $good_partition $volume
>     vos release $volume -verbose
> ?
>
> Your question prompts consideration of how one might write
> a script "unload_failing_disk" to get all the data off onto
> a free disk.  Move the unreplicated RWs first. But smallest
> volumes first or largest?
>

I tend to move the big ones first. That way, it's less chance of filling up the
other partitions such that the big ones won't fit anywhere.

>
>
> We recently started using a file server with RAID arrays
> of SSA disks in an external drawer. The SSA adapter
> in the IBM RISC System/6000 fileserver is configured to use
> a "hot standby disk". When a disk failure is detected, the
> SSA adapter automatically takes care of copying data from
> a failing disk to the "hot standby". This is all "invisible"
> to the AFS fileserver process because there is another layer
> of abstraction between "hdiskN" and the real SSA disk.
>

That's Gold for a SysAdmin. If you can afford it or persuade someone to pay for
it, then get hot spares.
Even if AFS makes such situations acceptable, it's an order of a magnitude less
work just to find an email telling you that "a disk was broken, but there were
no data loss, please replace the bad disk when you have some spare minutes"

---
Christer Bern�rus
Chips (DCE) Project Manager
Chalmers University of Technology
SE-412 96 G�teborg, Sweden

Voice: +46 (0)31 772 8656
Fax: +46 (0)31 772 8660
WWW: http://www.cs.chalmers.se/~bernerus
E-mail: [EMAIL PROTECTED]


Reply via email to